BeamCaster Overview

This experimental project implements a CPU-Side voxel raycasting technique using a beam-based acceleration method. Instead of casting a ray for every pixel, it processes the scene in 8x8 pixel blocks and dynamically adjusts traversal speed and precision after each hit.

A Beam here consists of two parts: a coherent group of rays (the "witness rays") that drives the global advancement within the 8×8×LOD block, and a second part in which the full set of rays for the 8×8 block is advanced to scan the space, starting from the point where the witness beam stopped.

Beam Traversal: Step-by-Step

• The Beam Caster groups rays into beams to improve efficiency.
• Witness rays are used within each block (defined with a binary mask) to detect voxels, minimizing check counts.

The algorithm operates in three main phases:

1. Fast traversal phase
   Witness rays advance at maximum speed, 8x8 voxels per step (coarse LOD).
   If no witness ray detects a voxel, the entire 8x8 block is skipped, and fast traversal continues.
   This repeats until the first hit is detected.

2. LOD switching phase after the first hit
   When a witness ray detects a voxel, the algorithm switches to a higher level of detail.
   It rolls back slightly and resumes the beam's progression at the new LOD.

3. Scanning precision at the finest level
   When a unit-sized voxel is hit, the algorithm switches to unit precision mode.
   It resumes from where the accelerated beam previously stopped and scans voxel-by-voxel, processing the full 8x8 block with all rays.

Loading Voxel Regions

Voxel Editor

To create voxel regions for this project, you can use the following editor:
https://github.com/mcidclan/voxelander-voxel-editor

Blender Script

Alternatively, use the Blender 3.x.x Python script available in the tool folder to export voxel regions from a voxelized mesh.

File Format

You can export multiple voxel files to be loaded by the renderer. Name them sequentially:

object\_0.bin
object\_1.bin
object\_2.bin
etc.

Configuration

You can adjust and experiment with the Beamcaster configuration using the opt.h file located in the /bmc folder.

This file centralizes key parameters such as:

• Rendering options: toggle FPS display, fog effects, and visual distortions (barrel, spherical, lens).
• Camera settings: origin scale, step size, near plane distance, default position and angles.
• Space and frame size: 3D space dimensions, buffer sizes, and frame scaling factors.
• Raycasting behavior: ray step factor, min step, max depth, base cut, and color depth factor.
• Beam parameters: ray count per beam, bitmask, origin detection, and boost level.
• 2D rendering layout: grid size, scissor bounds, and vertical displacement.

You can change these constants to explore different visual or performance profiles.

m-c/d